Final version: May 6, 1994 Trading Blocs and the Americas: The Natural, the Unnatural, and the Super-Natural Jeffrey Frankel, Ernesto Stein and Shang-jin Wei This paper was written for the Sixth Inter American Seminar in Economics, organized by Sebastian Edwards and Gustavo Marquez, sponsored by the National Bureau of Economic Research, Cambridge, MA., and held in Caracas, Venezuela, May 28-29, 1993. The full version of the paper appears as a 1994 U.C. Berkeley CIDER Working Paper, No. C94-034. This shortened version is to be published in an issue of the Journal of Development Economics, edited by Sebastian Edwards. The authors would like to thank Benjamin Chui and Xiong Bai Fan for research assistance, and Warwick McKibbin, Gary Saxonhouse and Alan Winters, for supplying data. They would also like to thank for support the Center for International and Development Economics Research, funded at U.C. Berkeley by the Ford Foundation, and the Japan-United States Friendship Commission, a U.S. government agency. Jeffrey Frankel is Professor of Economics, University of California, Berkeley and Senior Fellow at the Institute for International Economics, Washington DC Ernesto Stein is Economist, Inter-American Development Bank Shang-Jin Wei is Assistant Professor, Kennedy School, Harvard University JEL classification no.: F15 keywords: Free Trade Area, trading blocs, MFN, gravity model, regional trade preferences, MERCOSUR, Andean Pact, NAFTA

Trading Blocs and the Americas: The Natural, the Unnatural, and the Super-Natural Jeffrey Frankel, Ernesto Stein and Shang-jin Wei Summary Is world trade becoming more regionalized, as a result of preferential arrangements such as NAFTA, the Andean Pact and MERCOSUR? If so, is this deviation from the principle of MFN (non-discriminatory trade policies) good or bad? This paper attempts to answer both questions. Using the gravity model to examine bilateral trade patterns throughout the world, we find evidence of trading blocs in the Western Hemisphere and elsewhere, as in earlier work. Intra-regional trade is greater than could be explained by natural determinants: the proximity of a pair of countries, their sizes and GNP/capitas, and whether they share a common border or a common language. Within the Western Hemisphere, MERCOSUR and the Andean Pact countries appear to function as significantly independent trading areas, but NAFTA much less so (as of 1990). The intra- regional trade bias within MERCOSUR increased the most rapidly during the 1980s. In East Asia, on the other hand, increased intra-regional trade can be explained entirely by the rapid growth of the economies. We then turn from the econometrics to an analysis of economic welfare. Krugman has supplied an argument against a world of three trading blocs (that they would be protectionist), in a model that assumes no transport costs. He has supplied another argument in favor of trading blocs, provided the blocs are drawn along "natural" geographic lines, in a model that assumes prohibitively high transportation costs between continents. In this paper we attempt to resolve the Krugman vs. Krugman debate. We complete the model of the welfare implications of trading blocs for the realistic case where inter-continental transport costs are neither so high as to be prohibitive nor zero. We consider three applications of the model. (1) Continental Free Trade Areas (FTAs). We show that it is not only Krugman's "unnatural" FTAs that can leave everyone worse off than under MFN, but that under conditions of relatively low inter-continental transport costs, FTAs that are formed along natural continental lines can do so as well. We call such welfare-reducing blocs super-natural. (2) Partial regionalization. We find that partial liberalization within a regional Preferential Trading Arrangement (PTA) is better than 100 per cent liberalization. The super-natural zone, where the regional trading arrangement, in contrast to the Article 24 provision of the GATT reduces welfare, occurs for combinations of low inter-continental transport costs and high intra-bloc preferences, i.e., when the regionalization of trade policy exceeds what is justified by natural factors. (3) The formation of several sub-regional PTAs on each continent. We find that multiple FTAs on each continent could lower welfare, but that multiple PTAs, with partial internal liberalization, would raise welfare. We conclude the paper with an attempt to extract estimates of transportation costs from the statistics. Estimates suggest that trading blocs on the order of the EC are in fact super-natural. 4/94

1

We have recently seen an upsurge around the world of steps toward Free Trade Areas

and other special regional trading arrangements, from the European Union (EU, formerly the

EC) to the Association of SouthEast Asian Nations (ASEAN). Currently the momentum for

regionalization appears as strong in the Western Hemisphere as anywhere.

Most of the regional trade agreements that were announced in the past did not initially

come to much, such as the 1960 Central American Common Market (CACM), the 1960 Latin

America Free Trade Association (LAFTA), or the 1969 Andean Pact.1 More recent

agreements have been more serious, however. The Canadian-U.S. Free Trade Agreement was

successfully concluded in 1988, and went into effect in 1989. MERCOSUR was negotiated

between Brazil, Argentina, Uruguay and Paraguay in March 1990, scheduling an elimination of

all regional tariffs by the end of 1994 (though it is likely to run a bit behind the planned time-

table). Venezuela and Colombia reinvigorated the Andes Pact in November 1991, establishing

a common market by the end of 1992. More agreements are in the works, throughout the

Western Hemisphere2 and elsewhere.

In the 1990s, the talk has moved to expansion of the regional trading arrangements.

The North American Free Trade Agreement (NAFTA), was negotiated between the U.S.,

Mexican and Canadian governments in 1992, and went into effect January 1, 1994. There are

provisions to add other Western Hemisphere countries; the Clinton Administration has

confirmed that Chile is the first in line to negotiate joining NAFTA. Brazil in March 1994

proposed extending MERCOSUR into a customs union spanning all of South America, whether

as a rival bloc to North America or as a step toward hemisphere-wide integration.3 In Europe,

four members of the European Free Trade Association (EFTA) will join the EU, if their

populations approve in 1994 votes. In Asia, the Malaysians would like to expand ASEAN into

an East Asian Economic Caucus.

Table 1 presents statistics on the intra-regional share of trade undertaken by members of

2

these groupings. Intra-regional shares increased between 1965 and 1990 in some parts of the

world: from 0.8 per cent to 2.6 per cent among the Andean countries, from 35.8 per cent to

47.1 per cent among the EC 12, and from 19.9 per cent to 29.3 per cent among the East Asian

countries.

All this regional activity leaves some observers concerned that world trade is becoming

more regionalized. Is the apparent movement toward regionalization of the world trading system

good or bad? Let us begin by reminding ourselves that such a question is an exercise in the

"second best."

First-best would be a worldwide regime of free trade, where all countries agree to

refrain from erecting barriers and there is a serious international institution to enforce the

agreement. Modern trade theory, with its emphasis on imperfect competition and so on, has

done little to change this bottom line.4 But the first-best is an ideal that is rather unlikely to be

reached in practice. What sort of international trading arrangement is second-best?

Since its founding, the GATT has been predicated on the assumption that second-best is

a regime where each member accords others the status of Most-Favored Nation (MFN), i.e.,

treats its trading partners equally. The GATT incorporated an important exception to the MFN

principle in its Article 24: a subset of members could form a Free Trade Area (FTA), provided

certain conditions were met, including that barriers within the FTA were removed completely,

rather than only partially, and that barriers against non-members not be raised.

Arguments for the merits of the MFN-cum-Article 24 system could take either of two

possible tacks.5 First one might try to argue, in a static economic sense, that the formation of

FTAs under the conditions specified in Article 24 is likely to raise economic welfare, and that

other deviations from the MFN principle are not.6 Second one could argue, in a dynamic

political economy sense, that FTAs can act as stepping stones, which help build the political

support necessary to negotiate freer trade worldwide.7 Neither of these possible arguments is

3

especially clear or well-established. It is the first that we examine critically in this paper.

Paul Krugman has helped to focus the recent debate on whether a global trend toward

the formation of trading blocs would be a good thing or a bad thing. But he has supplied

equally clever arguments on both sides. In his first contribution (Krugman, 1991a), he focused

on the idea that when individual countries form larger groupings, they are liable to become

more protectionist, and thus to move farther from the ideal of world free trade. The reasoning

was that as a group they would set higher tariff levels vis-a-vis the rest of the world, since they

would have more monopoly power to exploit. Units were assumed to set tariffs at a self-

maximizing optimal level.8 He showed that world welfare is lower with a few trading blocs

than with the extremes of one or many, and that for specific plausible parameter values, three

turned out to be the worst possible number of blocs to have!

His second contribution, Krugman (1991b), provided a useful review of the whole array

of issues and factors involved. But it also included a very simple argument that leads to the

diametrically opposite conclusion from the first one, that trading blocs are good. It is observed

that even without the formation of regional free trade areas or preferential trading arrangements

of any sort, countries trade more with their neighbors than with countries from which they are

far removed, in part because of transportation costs. To the extent that there is less inter-

continental trade, there is less to be diverted (in the language of classical customs union theory).

Imagine, in the limit, that transoceanic transportation costs were so high that all trade

took place within continents. Then it must follow from standard trade theory that removal of

trade barriers within each continent, that is, the formation of regional free trade areas, would be

a good thing: this move within each area would represent the first-best solution of free trade

within its own relevant world. Krugman's conclusion is that, to the extent that trade follows the

"natural" lines dictated by proximity, the formation of regional trading blocs is good. Such

natural blocs are contrasted with "unnatural blocs", free trade agreements between individual

4

countries on different continents, which are less likely to be welfare-improving.9

Each of these two arguments is valid within its own assumptions. One way to

characterize them is as the limiting polar cases of zero inter-continental transportation costs and

infinite inter-continental transportation costs, respectively. The analysis, to be complete, cries

out for a more general model that can handle the intermediate realistic case where transportation

costs between continents are less than infinite, while greater than zero (and greater than

transportation costs within continents).

One can imagine several possible rules regarding general preferential trading

arrangements (PTAs), in addition to the question of whether the FTA deviation from MFN

practice should be encouraged or allowed at all. First, should FTAs be restricted to natural

trading partners, as Krugman (1991b) suggests? This would mean that FTAs could only be

formed among countries that are located in the same part of the world (e.g., the Western

Hemisphere) or perhaps only among neighbors located in the same sub-region (e.g., North

America, which would exclude even an agreement between NAFTA and Chile). Second, is

the rule sensible that technically requires 100 per cent liberalization within a grouping, i.e., that

allows only FTAs? Or should partial liberalization be allowed, as de facto prevails in most

PTAs? Is there an optimal degree of regionalization that should be encouraged? Third, is it

desirable to allow the formation of sub-continental trading arrangements like the Andes Pact or

MERCOSUR?

We attempt to do several things in this paper. First, we shall measure the extent to

which regionalization is actually taking place, by looking at the magnitude of bilateral trade

flows after one adjusts, by means of the gravity model, for such natural determinants of bilateral

trade as GNPs and proximity.

We consider two alternative possibilities regarding the relevant place to draw the

boundaries of the regional groupings: at the level of continental blocs, or at the level of sub-

5

continental FTAs consisting of a few members each (e.g., NAFTA, MERCOSUR, and the

Andean Pact).10

That the share of intra-regional trade is increasing within a given grouping, as in Table

1, does not necessarily mean that the members of this grouping are undertaking explicit

discriminatory trade policy measures to bring this about. Rapid growth in intra-regional trade

could be the result of natural factors, i.e., rapid growth in per capita GNPs. Indeed we find

that this is the case for East Asia. In Europe and the Americas, on the other hand, there

appears to be a statistically significant role for regional trade policies, even after correcting for

natural determinants.

Second, we address welfare implications of different possible rules for the formation of

preferential trade groupings. At a theoretical level, we shall attempt to complete the Krugman

model of the welfare implications of trading blocs for the realistic case where transportation

costs between continents are neither so high as to be prohibitive nor so low as to be the same as

costs among neighbors. We consider three applications of the model in turn.

We start with continental FTAs, where intra-continental tariffs are completely

eliminated. We shall see that it is not only unnatural FTAs that can leave everyone worse off

than under MFN, but that under certain conditions FTAs that are formed along natural intra-

continental lines can do so as well. We call such welfare-reducing blocs super-natural.11 We

shall see in simulations that this possibility may obtain, in particular, when intercontinental

transportation costs, while not necessarily as low as intracontinental costs, are as low as 10 or

20 per cent.

Next we consider two different kinds of "partial regionalization." One is partial

preferential treatment within regional trade groupings. We allow for the formation of

Preferential Trade Agreements (PTAs) that differ from the FTAs in that the tariff level is

reduced among partners, but not necessarily eliminated. Even though it is technically prohibited

6

by Article 24, many existing regional arrangements are in fact of this partial kind.12 We will

show that a partial movement towards regional integration, as in the case of PTAs with

preference below 100%, is superior to FTAs. The super-natural zone, where the regional

trading arrangement reduces welfare, occurs for combinations of low intercontinental transport

costs and high intra-bloc preferences.

A different way to look at a partial trend toward regionalization is to recognize that each

continent has many countries, and to consider the formation of several sub-blocs within each

continent. The final application of the model is to the question of sub-regional FTAs. We have

in mind, for example, the regionalization of trade within the Americas into four FTAs

consisting of NAFTA, Central America, the Andean Pact and MERCOSUR. We find that

such an arrangement, like continental FTAs, would be worse than the status quo of MFN. If

the constraint of Article 24 is relaxed however, and partial liberalization within each regional

trading arrangement is allowed, then the formation of several PTAs within each continent is a

good thing, although continent-wide PTAs are even better.

In the final part of the paper, we attempt to get a better idea of which of the theoretical

welfare possibilities is actually most likely in practice by adopting estimates of the parameters

from the 1965-1990 data on bilateral trade that are used in the first part of the paper. An

estimate of intra-continental transport costs based on the ratio of c.i.f. to f.o.b. values (and trade

shares) is 10 per cent, suggesting that super-natural blocs are a real danger. Distance may

generate costs beyond the freight and insurance required by physical transport of goods

however. An alternative to the c.i.f./f.o.b. calculation is to use the gravity model estimates of

the effect of distance on trade. A tentative estimate of the intra-continental parameter, based on

the gravity model, is 16 per cent. Such an estimate, combined with our other simulation

parameter values, would imply that negative returns to regionalization may begin to set in when

regional preferences reach about 23 per cent.

7

Most of our conclusions regarding economic welfare presume worldwide symmetry. In

other words, we look at the consequences of a worldwide regime of allowing continental blocs

or regional FTAs to form; the consequences of the unilateral formation of a single bloc or FTA

in one part of the world is not addressed in this paper.13 It should be noted from the outset that

many of the conclusions are tentative, and that many possible considerations are omitted from

the analysis.

2. Are Regional Trade Blocs Forming?

Frankel (1993) applied to the trading bloc question the natural framework for studying

bilateral trade, the gravity model. The gravity model says that trade between two countries is

proportionate to the product of their GNPs and inversely related to the distance between them,

by analogy to the formula for gravitational attraction between two masses. It has a fairly long

history and fits the data remarkably well empirically, though its theoretical foundations are

limited.14 There are not many recent applications of the gravity model to a large cross-section

of countries throughout the world. Three others are Wang and Winters (1991), Hamilton and

Winters (1992), and Havrylyshyn and Pritchett (1991).15

Frankel (1993) and Frankel and Wei (1994), looking at the period 1980-1990, found

that: (1) there are indeed intra-regional trade biases in the EC and the Western Hemisphere,

and perhaps in East Asia; but (2) the greatest intra-regional bias was in none of these three, but

in the APEC grouping, which includes the U.S. and Canada with the Pacific countries; and (3)

the bias in the East Asia and Pacific groupings did not increase in the 1980s, contrary to the

impression that many have drawn from intra-regional trade statistics such as are reported in

Table 1.

Frankel and Wei (1993a, 1993b) extend those results in a number of directions. The

papers consider various econometric extensions of the original gravity model estimation: the

8

inclusion of pairs of countries that are reported as undertaking zero trade, and a correction for

heteroscedasticity based on the size of the countries. The time period is extended 15 years

farther back. The results turn out to be robust to these extensions. The papers also considered

some economic extensions, in particular testing whether stabilization of bilateral exchange rates

has been a factor in promoting intra-regional trade.

Here we focus more on sub-regional groupings in a world of many countries, especially

Western Hemisphere groupings such as NAFTA, MERCOSUR, and the Andean Pact. We

also describe economic extensions, such as the effects of factor-endowment differences and the

difference between a customs union effect and a PTA effect, and social/political extensions such

as the effect of common languages or colonial links.

2.1 The Gravity Model of Bilateral Trade

One cannot meaningfully investigate the extent to which regional policy initiatives are

influencing trade patterns without holding constant for natural economic determinants. The

gravity model offers a systematic framework for measuring what patterns of bilateral trade are

normal around the world. A dummy variable can then be added to represent when both

countries in a given pair belong to the same regional grouping. The goal, again, is to see how

much of the high level of trade within each region can be explained by simple economic factors

common to bilateral trade throughout the world, and how much is left over to be attributed to a

special regional effect.

The dependent variable is trade (exports plus imports), in log form, between pairs of

countries in a given year. We have 63 countries in our data set, so that there are 1,953 data

points (=63x62/2) for a given year.16

The two most important factors in explaining bilateral trade flows are the geographical

distance between the two countries, and their economic size. Indeed, these two variables give

the gravity model its name.

9

A large part of the apparent bias toward intra-regional trade is certainly due to simple

geographical proximity. Indeed Krugman (1991b) suggests that most of it may be due to

proximity, so that the three trading blocs are welfare-improving "natural" groupings. Despite

the obvious importance of distance and transportation costs in determining the volume of trade,

empirical studies surprisingly often neglect to measure this factor. Our measure is the log of

distance between the two major cities (usually the capital) of the respective countries. We also

add a dummy "Adjacent" variable to indicate when two countries share a common land border.

Entering GNPs in product form is empirically well-established in bilateral trade

regressions. It can be justified by the modern theory of trade under imperfect competition. In

addition there is reason to believe that GNP per capita has a positive effect on trade, for a given

size: as countries become more developed, they tend to specialize more and to trade more.17

The equation to be estimated, in its most basic form, is:

(1) 0.

EA, EC, and NAFTA are three of the dummy variables we use when testing the effects of

membership in a common regional grouping standing for East Asia, European Community, and

North America.

Table 2 reports results that extend from 1965 to 1990. We find all four standard gravity

variables to be highly significant statistically (> 99% level).

The 1990 coefficient on the log of distance is about -.6, when the adjacency variable

(which is also highly significant statistically) is included at the same time. This means that when

the distance between two non-adjacent countries is higher by 1 per cent, the trade between them

falls by about .6 per cent. [We have also tried distance measures that take into account the

10

greater distances involved in sea voyages around obstacles like the Cape of Good Hope and

Cape Horn, generously supplied by Winters and Wang (1992), with little effect on the results.]

The coefficient of distance varies a bit over the course of the earlier observations, but

with no clear trend. Disaggregated results show higher distance effects for manufactures than

for agricultural products or other raw materials, which are bulkier.18 These findings suggest to

us that physical transport costs may not be the most important component of costs associated

with distance.

The estimated coefficient on the product of per capita GNPs varies in the .26-.40 range

from 1965 to 1980, indicating that richer countries do indeed trade more. This term declines

during the 1980s. The estimated coefficient for the log of the product of the two countries'

GNPs holds roughly steady at about .7, indicating that, though trade increases with size, it

increases less-than-proportionately (holding GNP per capita constant). This reflects the familiar

pattern that small economies tend to be more dependent on international trade than larger, more

diversified, economies.

2.2 Estimation of trade-bloc effects

If there were nothing to the notion of trading blocs, then these four basic variables might

soak up all the explanatory power. There would be nothing left to attribute to a dummy

variable representing whether two trading partners are both located in the same region. In this

case the level and trend in intra-regional trade would be due solely to the proximity of the

countries, and to their rates of overall economic growth.

But we found that dummy variables for intra-regional trade are highly significant

statistically.19

In earlier results, if two countries were both located in the Western Hemisphere, they

traded with each other by an estimated 86 per cent more in 1980 than they would have

otherwise [exp(.62) = 1.86], after taking into account distance and the other gravity variables.

11

(Table 4 below shows even bigger effects.) Table 2 replaces the Western Hemisphere bloc

variable with separate dummy variables for three sub-regions: NAFTA, MERCOSUR, and the

Andean Pact. Tight standard errors and significant coefficients are not to be expected, in light

of the small number of observations: 3 (=3x2/2) for NAFTA, 6 (=4x3/2) for MERCOSUR,

and 10 (=5x4/2) for the Andean Pact. But the point estimates are of interest nonetheless, as

these are the groupings with explicit trade preferences. The estimates for MERCOSUR and the

Andean Pact turn positive in 1970, the latter significantly so in 1975 and 1980, and both

significant in 1990. Remarkably, members of MERCOSUR in 1990 trade with each other

eight times as much [exp(2.09)=8.08] as would similar neighbors elsewhere in the world. The

NAFTA coefficient only turns positive in 1985. As one would certainly expect from the

extremely small number of observations, it is not statistically significant.

As recently as 1980, the EC bloc effect was not statistically significant. The effect in

1985 is highly significant. A 1985 coefficient of 1.14 suggests that if two countries are both

located in the European Community, their bilateral trade is three times as high as it would

otherwise be [exp(1.14) = 3.13]. EFTA is never significant.

As in earlier results, the coefficient for the East Asian grouping [not including Australia

and New Zealand] is highly significant, but diminishes in the 1980s, rather than increasing as

often assumed. The same is true of APEC. The rapid growth of East Asian economies is in

itself sufficient to explain the increase in the intra-regional trade shares evident in Table 1.

We tried dropping the specific sub-regional groupings [in the Western Hemisphere] and

replacing them with one dummy variable to indicate whenever a pair of countries belongs to the

same PTA or FTA, regardless which one it is, and another to indicate whenever the pair

belongs to the same Customs Union or Common Market. The distinction is that in the latter

two arrangements, external tariffs are made uniform. The PTA/FTA variable is often

statistically significant, particularly when the tests are run on manufacturing products alone.

12

The CU/CM variable is not.20

Next, we added a dummy variable to represent when both countries of a pair spoke a

common language or had colonial links earlier in the century. We allowed for English,

Spanish, Chinese, Arabic, French, German, Japanese, Dutch, and Portuguese.[21] Two

countries sharing linguistic/colonial links tend to trade roughly 65 per cent more than they

would otherwise [exp(.5)=1.65].22 We tested whether some of the major languages were

more important than the others. Chinese is the only one that might qualify. [Two Chinese-

speaking countries appear to trade four times as much as other countries.23]

Somewhat surprisingly, the inclusion of the linguistic/colonial terms has little effect on

the other coefficients. The trade blocs remain significant, with increasing trends over the period

1965-1990 in each case but APEC and East Asia.

Finally, we also tried to capture classic Heckscher-Ohlin effects. First we tried

including bilateral absolute differences in GNP/capita figures for 1990, reported in Table 3. The

variable did not have the positive effect that one would expect if countries traded capital-

intensive products for unskilled-labor-intensive products. Rather, it had a moderately signficant

negative effect, as in the Linder hypothesis that similar countries trade more than dissimilar

ones.

Next we tried, in Table 4, gravity estimates that include more direct measures of factor

endowments: the two countries' differences in capital/labor ratios, educational attainment levels,

and land/labor ratios. The data (for a subset of 656 of our 1,953 pairs of countries) was

generously supplied by Gary Saxonhouse. There is only a bit of support for these terms,

specifically for capital/labor ratios and educational attainment in 1980. The coefficients on the

bloc variables and other effects change little qualitatively.

The gravity model results thus show that statistically significant regional trading

arrangements are indeed springing up in a number of places. The next question is whether this

13

trend constitutes an undesirable threat to the world trading system.

3. The Theory of Trade with Imperfect Substitutes and Transportation Costs

This and the next parts of the paper attempt to settle the Krugman vs. Krugman

controversy regarding the desirability of trading blocs, by constructing a more general model

that can handle the intermediate realistic case where transportation costs between continents are

less than infinite, while greater than zero (and greater than transportation costs within

continents). The ultimate goal is a preliminary matching-up of the theory up with the preceding

section's empirical estimates of the effects of transportation costs and regional trading

arrangements on the volume of bilateral trade, in order to allow an evaluation of different trade

arrangements.

3.1 The Differentiated Products Model

We work with a familiar model of trade under monopolistic competition due to Krugman

(1980). Our contribution is to extend this model to many countries, allowing for tariffs and

transportation costs, both within continents and between continents, and to apply it to study the

welfare implications of the formation of trading blocs.

The Krugman market structure has the property of ruling out strategic interaction among

firms. Goods enter symmetrically into the utility function

where ci is the consumption of the ith

variety. There is a large number of

goods being produced (n), but this

number is much smaller than the

potential number of goods or varieties.

This utility function results in preference for variety by the consumers. The higher the

parameter è, the lower the love for variety. In the limit of perfect substitutability, è=1:

0

14

consumers care only about the total quantity consumed, not about different brand names. In the

limit of complete love for variety, è=0: consumers, like stamp collectors or bird watchers,

care only about the number of varieties consumed, and not at all about the quantity of each.

Labor is the only factor of production. The total national supply of labor is 0.

Increasing returns are introduced by assuming a fixed cost and

a constant marginal cost in the production of each of the

varieties. We assume that individual consumers maximize their

utility, individual firms maximize their profits, and free

entry assures a zero-profit equilibrium. Under these simple

assumptions, the scale of output of each variety does not

depend on the size of the economy. Rather, it is the number

of varieties n that increases when the size of the economy (L)

increases:

where 0 is a parameter representing the

fixed costs of setting up production of a

new variety. Notice that in the extreme

special case of zero substitutability

( 0=0), the bare minimum (one unit) of each of L/ 0

varieties will be produced, since consumers care only about

the number of varieties available. (Details of this

derivation, and of others below, are given in Stein and

Frankel, 1993.)

To see the gains from international trade that arise here from the opportunity to consume

a greater variety of goods, we assume that countries have similar tastes and technologies. If we

0

15

have two countries of equal size, allowing for unfettered trade will double the number of

available varieties in each country and thus raise utility.

3.2 Introduction of Transport Costs and Tariffs

We will think of the world as being divided into a number of continents (C), each of

them equidistant from one another. Each of these continents is composed of a number of

countries (N). Transport costs will be assumed, following Krugman (1980), to be of

Samuelson's iceberg type, which means that only a fraction of the good shipped arrives; the rest

is lost along the way.24

The cost of transport within a continent will be represented as a, while that of transport

significant (across the ocean), is an additional b, where 0#a,b#1. The fraction of a good

shipped intra-continentally that arrives to the market is 1-a. The fraction of a good that arrives

in the case of trade between countries in different continents is (1-a)(1-b).

Tariffs will be treated in a standard way. When a consumer buys a foreign good, the

government levies an ad-valorem tariff t. [Our basic theoretical model will assume that the

tariff is levied as a percent of the value of the good expressed in f.o.b. terms, i.e., not including

transportation costs. For some purposes it may be more convenient, as well as more realistic,

to assume that it is levied as a proportion of the value of the good in c.i.f. terms, i.e., including

transportation costs.25] The level of tariffs is exogenous, and assumed to be uniform across

countries, representing the MFN principle, until we are ready to examine preferential trading

arrangements.

For simplicity, we will assume that each one of the countries is equal in size. The

symmetry of the model now assures that the producers' prices are the same in every country, as

well as the number of varieties and the quantity of each variety produced in every country.

Prices of foreign goods faced by home consumers are higher than prices of home goods, due to

transportation costs and tariffs. If the producer prices in every country are p, then the price the

16

domestic consumer will have to pay for every unit of foreign good consumed will be:

where the subscript c refers to goods imported from within the continent, and nc inter-

continentally. Notice that import prices depend positively on tariffs and transportation costs. In

the absence of tariffs, the prices faced by the home consumers are pc=p/(1-a) and pnc=p/(1-

a)(1-b).

Since the home consumer will be paying different prices for the consumption of home

and foreign products, he or she will be consuming them in different quantities. The next step is

to derive, from the utility function, the consumption of each foreign variety (both from neighbor

countries and from countries in other continents) relative to the consumption of each home

variety. We begin by assuming that tariffs t are levied.

The first order conditions for the consumer's problem yield

the relative consumption of each variety:

where cic and ci

nc are the domestic consumer's consumption

of foreign varieties, from countries within the continent and

across the ocean respectively, and cih is the domestic

consumer's consumption of the home varieties. Thus the

elasticities of demand are åx=1/(1-è).

We derive the bilateral volume of trade (BVT) in Frankel, Stein and Wei (1993).26 It

has the desired property that, to the extent that tariffs and transportation costs raise the price in

an importing country, the volume of trade will fall, to an extent determined by the elasticity.

By assigning values to the parameters a, b, t, è, N and C, we can obtain the exact effect on

0

0

17

BVT of any symmetrical arrangement.

In order to explore the desirability of potential trading blocs, we now need to introduce

a measure of welfare.

3.3 Welfare implications of trade agreements

Given that we are working with a symmetric model, the natural way to look at world

welfare is to derive the utility of a representative individual in any country. To determine the

utility of the consumer, we need to know how much he or she is consuming of each good, and

introduce these values into the utility function. Equations (5) and (6) above gives us the relative

consumption of each home and foreign variety, so we only need to determine the consumption

of each home variety, cih. We do this by expressing the budget constraint in terms of ci

h, and

taking into account the redistribution of the tariff revenue to consumers.

If we normalize n, p, and w to be 1, we can obtain, after some algebra

Once we have the consumption of domestic varieties, the consumption of foreign varieties can

be obtained from the relative consumption equations (5) and (6):

0

0

18

Replacing these into the utility function, we obtain the value of the utility of the representative

individual:

Given values for the parameters a, b, t, è, N and C, we can plug the price equations (4) into

(7), plug the value of cih into equation (8), and finally substitute into (9) to obtain the value of

the utility of the representative individual, which is our measure of world welfare.

Equation (9) is the expression for utility in the absence of free trade agreements. It is

simple to calculate utility under other arrangements in the same manner. When trading blocs are

formed, we just introduce the new set of relative prices faced by the home consumers into the

model, and we can obtain new results for utility in a similar way.

4. Welfare Implications of Free Trade Areas

We have presented a model that allows us to analyze the desirability of different trade

arrangements from the perspective of world welfare. We now present some applications.

4.1: The number of blocs and welfare in the absence of transportation costs

The purpose of this exercise is to check that our model yields Krugman's U-shaped

welfare curve as a function of the number of blocs, in the absence of transportation costs. We

assume a world consisting of 60 countries, and study the welfare implications of dividing the

trading system equally into different numbers of blocs. Figure 1 shows the results of our

simulations for a value of è = 0.75, and tariffs of ten, twenty and thirty percent.27 We can see

that welfare is minimized for a small number of blocs, three in the cases of twenty and thirty

percent tariffs, and two blocs in the case of a tariff rate of ten percent. Welfare increases

0

19

gradually beyond the minimum-welfare number of blocs.28

In Krugman's model (1991a), there are two reasons for the increase in welfare as the

number of blocs becomes larger. One reason is that blocs set tariffs optimally, and become less

protectionist as the market power of each one declines. The other reason is that as the number

of blocs increases, a larger portion of their demand is satisfied from outside the bloc, and tariffs

become less distortionary. Tariffs introduce a wedge between the prices of bloc varieties and

those of non-bloc varieties, but not between two non-bloc varieties. The greater the number of

non-bloc varieties relative to those from within the bloc, the smaller the distortionary effect of a

given tariff level. In our model, where tariffs are assumed exogenous, the shape of the curve is

explained completely by this latter reason.29

4.2: Transport costs and the welfare effects of continental free trade agreements

In this application, we study how the effect of the formation of free trade agreements on

welfare depends on intercontinental transportation costs. Thus we are able to fill in the realistic

intermediate case between Krugman's polar cases of zero and infinite intercontinental

transportation costs.

For the purposes of the simulations presented here, transportation costs within continents

are for simplicity kept at zero. Our base-case substitution parameter is è=0.75 and our base-

case worldwide level of multilateral tariffs is t=0.3. We begin with a Krugman (1991b) world

that consists of three continents, with two countries in each continent.

Figure 2 shows the percentage change in welfare associated with the formation of

trading blocs, both of the natural and unnatural type. There is a critical level of intercontinental

transportation costs b, that governs the welfare effects. For the case of natural trading blocs,

where each country forms a bloc with its neighbor, the critical value of b is approximately

0.15. For higher values of b, the formation of continental trading blocs will result in

improvements in welfare. (Remember, in the limit, Krugman's case where b=1.) For lower

20

values of b, continental blocs would reduce welfare. (Remember the limit case where b=0.)

As noted in the introduction, we label such welfare-reducing arrangements "super-natural

blocs", to indicate that intercontinental transportation costs are not high enough to justify the

formation of blocs even along the lines of geographical proximity. The benefit of forming

trading blocs becomes much larger, and the danger of entering the super-natural zone

diminishes, as t and è increase. (An appendix to CIDER Working Paper No. C94-034

undertakes the sensitivity analysis.30)

Unnatural trading blocs, where each country forms a bloc with one other country

outside the continent, result in distinctly lower welfare for small values of b. When b=0 they

coincide with natural blocs. [Unnatural blocs then have a steadily smaller effect as b tends to 1.

The reason for this is intuitive: as b gets closer to 1, the bilateral volume of trade between

countries in different continents will tend to zero, whether they belong to the same bloc or not.

Therefore, the formation of unnatural trading blocs has only negligible effect on welfare when

intercontinental transport costs are very high. The limit is the polar case of no intercontinental

trade.]

5. The Effects of Preferential Trading Arrangements with Partial Liberalization

Regionalization can fall short of full continental FTAs, either with respect to the

magnitude of the tariff preferences or the fraction of the continent covered by each bloc.

5.1: Allowing for Preferential Trade Agreements on each Continent

In this application, we will have another look at trading blocs of the "natural" kind

(among neighbors), but we will allow for the formation of PTAs, i.e., partial liberalization. To

do this, we need to modify our model slightly. The tariff level between partners, instead of

zero, will now be (1-k)t, where 0#k#1, and k is the degree of preference for intra-bloc trade

or intra-bloc liberalization. The price of partner varieties faced by domestic consumers now

21

becomes:

Until now we were only

considering the two special cases of

k=0 (absence of blocs) and k=1

(Free Trade Areas). Now intra-bloc

preferences can be set at any level. We will begin, as in the previous application, with a world

that consists of three continents, each formed by two countries.

What is the level of intra-bloc preference that will maximize welfare? Figure 3 shows

the welfare level as a function of k, for t=0.3, è=0.75, a=0, and several values of b.31 For

example, for b=0.1 the level of welfare for the extreme of full continental FTAs (k=1) is

lower than the opposite endpoint of MFN or non-discrimination (k=0). This is the case of the

super-natural FTA.

The important thing to notice in Figure 3 is that for every level of intercontinental

transport costs, the degree of intra-bloc preference associated with maximum welfare is in

between 0 and 1. This implies that, in general, PTAs with less than 100% preference are

superior to FTAs.32 This result is not new in the literature. It was first suggested by Meade

(1955). But it is significant if we contrast it with GATT's article 24, which allows for FTAs

and Customs Unions as exceptions to the Most Favored Nation (MFN) clause, but not for PTAs

with less than 100% preference.33

Figure 3 suggests that starting from the absence of trading blocs, a small movement in

the direction of increased regionalization (by increasing intra-bloc preference) is always a good

thing. We can say that there are positive returns to regionalization up to the point of maximum

welfare, and negative returns to regionalization thereafter.

Another way of looking at this issue is to show all possible combinations of

intercontinental transport cost b and intra bloc preference k [and for a given set of values

0

22

chosen for the other parameters]. Frankel, Stein and Wei (1993) does this for a world of six

countries (= 3 continents of 2 countries each) and tariffs levied in c.i.f. form.34

In reality, the world of course consists of more than three continents of two countries

each. In Figure 4 we here repeat the experiment for the more realistic (if still stylized) case

where the world consists of four continents of 16 countries each. We could get to four

continents either by counting North and South America separately, or adding the

Mideast/Africa. This 64-country set-up has the virtue of corresponding roughly to the data set

in our gravity model. The solid line represents the level of intra-bloc preference that maximizes

welfare at each level of transportation cost b. Below this line, there are positive returns to

regionalization, i.e., increasing the degree of preference will result in higher welfare. Above

this line, increases in the preference are welfare-reducing. We call this the area of negative

returns to regionalization NRR.

Within the NRR area, the dotted line represents, for every level of intercontinental

transportation cost, the intra-bloc preference level that yields the same welfare as k=0 (i.e., the

absence of trading blocs). The trade arrangements that lie above this dotted line are welfare-

reducing, relative to the status quo of MFN. These are the ones we call super-natural trading

blocs.35

We see that negative returns to regionalization set in sooner than in the six-country case.

If inter-continental transport costs are .2, then the world reaches the welfare optimum when

intra-bloc preferences are as low as 27 per cent, and enters the super-natural zone when they

are 51.5 per cent. If inter-continental transport costs are as low as .1, then negative returns to

regionalization set in even sooner.36

We have seen, within the terms or our model, that (1) a little bit of regionalization,

defined as a small degree of preferences for continental neighbors, is a good thing, but that (2)

100 per cent liberalization within a continental FTA is carrying regionalization too far. In terms

23

of static economic welfare, the apparent conclusion under our assumptions is that Article 24 of

the GATT would do well to relax the stipulation that liberalization within a preferential trading

area be complete.

5.2: Welfare Effects of Sub-continental Blocs

Application 3 has shown a sense in which a partial movement towards regionalization

may be better than a total one. We now look at another way in which "partial" trends toward

regionalization can be understood: the formation of multiple blocs on each continent. We have

in mind recent sub-continental FTAs of two countries each, like the Canada-U.S. FTA or the

customs union between Colombia and Venezuela that was instituted in January 1992. We also

wish to consider somewhat bigger groupings, like the NAFTA, CACM, MERCOSUR, and

Andean Pact in the Western Hemisphere.

For this purpose, we run a simulation where the world consists of 4 continents, each of

them containing 16 countries. This allows us to compare welfare under the MFN rule with that

associated with 8 sub-continental FTAs on each continent of 2 countries each, 4 of 4 each, or 2

of 8 each. The results of this simulation are seen in Figure 5. MFN in this figure is the starting

point: 16 sub-continental blocs, each formed by one country.

The Figure shows that the formation of FTAs between regional subsets of countries is

not a good idea, and the more countries that participate, the worse the idea. Even at the last

stage, when two half-continental blocs of 8 members each are merged into a continental FTA,

welfare falls slightly, if b is .2 or less.37 These results seem to bode ill for recent regional

agreements.38

We have found that the formation of a number of FTAs within each continent, for our

parameter values, lowers welfare regardless of the number and size of the FTAs. But we

found earlier that partial liberalization in continent-wide PTAs is better than both MFN and

fully-liberalized FTAs. Is the same true for the formation of a number of PTAs within each

24

continent?

Figure 6 addresses this question for the case where b=.2. The right edge confirms that

the formation of eight two-country FTAs on each continent reduces welfare, and larger blocs

are even worse. But for partial preferences, ideally in the range of 20 to 25 per cent, the

picture for multiple PTAs looks much better. Two-country PTAs are slightly better than the

MFN status quo (1-country groupings). Four-country PTAs are better still, and so on until the

optimum is reached at a continent-wide grouping of 16 countries (at which point preferences of

27 per cent are the precise optimum in the simulation, as we saw in Figure 4). In other words,

welfare increases monotonically with the size of the PTAs, rather than decreasing monotonically

as it did for the case of FTAs. Clearly the distortionary (or trade-diverting) effects are less

important when internal tariffs are only reduced partway. The pattern is similar when b=.1 or

b=.3, but the level of preferences that maximizes welfare for each size of PTA becomes

approximately 15 per cent and 25-35 per cent, respectively. [Graphs are available in CIDER

Working Paper C94-034, but omitted here to save space.]

Why might countries wish to negotiate small two-country PTAs that would raise welfare

only slightly, if larger PTAs would be even better? For the same reason that it seems to be

impossible to negotiate worldwide liberalization. Although these political economy

considerations lie outside the scope of our model, one could easily posit costs to international

negotiation that increase with the number of partners involved.39 We have in mind, not just the

hours, salaries, or airfares of the negotiators, but the adjustment costs of harmonizing standards

and administrative procedures or the difficulty of satisfying adversely affected interest groups.

Two-country PTAs could then be viewed as stepping stones or building-blocks for four-country

PTAs, leading to eight and, finally, to the continent-wide arrangement.

The stepping-stone idea would be particularly attractive if there was reason to think it

could be sustained across continents. We now turn attention to blocs formed across continents.

25

We consider a world formed by four continents and six countries per continent in order to

answer the following question: under what circumstances will it be beneficial for the world to

consolidate into two blocs, each formed by two continents?

For the parameters è=0.75 and t=0.3, we find that the consolidation will be beneficial

under any transportation costs a and b. We might have predicted this: if we look at Figure 1,

we can see that, in the absence of transport costs, two blocs are better than four for these

parameter values. And we know, from application 2, that consolidation is more likely to

improve welfare the higher the transportation costs.

The interesting cases are those that correspond to parameter values such that, in the

absence of transport costs, four blocs are better than two. In these cases, our results show that

there is a critical value of b above which the consolidation becomes beneficial. We ran several

simulations for è=0.6, different values of t (0.2 and 0.3) and different values of a (0, 0.2 and

0.5). We found that the critical value of b will be lower (and therefore consolidation more

likely to be improve welfare) the higher the tariff level and the higher the transportation cost a.

This kind of analysis can be useful to study the welfare consequences of potential blocs such as

the hemisphere-wide one, which would include both North and South America (here b is not

much larger than a on average), or a trans-Pacific grouping, as often discussed in meetings of

such organizations as APEC, PECC, and PAFTAD (where b is large).

6. Some Estimates of Transport Costs to Evaluate the Extent of Regionalization

To get a better idea where the world economic system lies in terms of the welfare spaces

mapped out above, we would like to have some estimates of the parameters, especially the

crucial magnitude of intra-continental transportation costs, b. We can think of four ways of

estimating b. First is direct data on bilateral shipping costs. The disadvantage here is that the

range of variation of actual shipping costs is extremely wide across modes of transport and

26

kinds of goods, especially as a percentage of value, and it would be difficult to know how to

aggregate different measures.

Second is the ratio of the c.i.f. value of a country's trade to its f.o.b. value. One

disadvantage here is that the data are not available on a bilateral basis. Another disadvantage of

using aggregate c.i.f./f.o.b. numbers is that they depend on the composition of trade (which is

in turn endogenous).

The ratio of total worldwide import values, including insurance and freight, to export

values is about 1.06.40 We can infer a rough upper bound on b by assuming that 6 per cent is

a weighted average of intra-continental costs and inter-continental costs:

.06 = ICS a + (1-ICS)(a+b-ab), or

b = (.06-a)/[(1-ICS)(1-a)] < .06/(1-ICS). (11)

We get our ICS estimate from Table 1. Considering only the set of 63 countries examined

statistically in the first part of the paper, the intra-continental trade share is about .4. Thus (11)

implies an upper bound on b of .06/(1-.4) = .10.

If 10 per cent is a realistic estimate of intra-continental transport costs, then we can see

from the simulations that super-natural trading blocs are a real danger. Indeed, for b=.10, our

base-case parameter values, and a world consisting of three continents of two countries each,

negative returns to regionalization set in when preferences are 52.4 per cent; any greater degree

of regional preference moves into the zone of negative returns to regionalization (Figure 3).

For this world, 95 per cent preferences are in the super-natural zone. For a world consisting of

four 16-country continents, negative returns set in even sooner (Figure 4). The optimum

degree of continental preferences is just over 16 per cent, and the super-natural zone begins at

32 per cent.

It is possible that the c.i.f./f.o.b. ratio substantially understates the costs of trade by

focusing solely on the cost of physical transport. Within the confines of our theoretical model,

27

the parameter b could be estimated in a simple way from the data on intra-regional trade shares,

if we were willing to assume that the observed current tendency for countries to trade with

neighbors was the result solely of geographical proximity, and not of preferential trading

policies.41 We pursue this logic next.

If b is high, then inter-continental trade will be low relative to continental trade. In

CIDER Working Paper C94-034, we derive an expression for b,

b = 1 - 0,

(12)

where 0 and 0 are the continental and inter-

continental ratios of demand, respectively (relative to home

varieties).

Total intra-continental trade on a continent is 0. Total trade

undertaken by the continent with other continents (including

both imports and exports) is 0.

Thus ICS = 0/[ 0+ 0]. In the special case

where intra-continental trade as a share of GNP in each

country i is the same, and the inter-continental share of each

country is the same, the intra-continental trade share becomes

ICS = 0/[ 0+ 0] =

28

0. (13)

It follows that

ICS = 0. Solving for 0 and substituting

into (12),

b = 1 - 0. (14)

The set of countries from which our trade data come can be approximately described as four

continents (including Africa/Mideast along with the other three42) consisting of 16 countries

each. Substituting C=4, N=16, and 0=.75, into equation (14), we

obtain an illustrative estimate of b = .383. This is quite a

high estimate of intra-continental costs, and it would imply a

corresponding reduction in the risk of trade policies becoming

excessively regionalized.

We know from our gravity estimation, however, that statistically significant tendencies

toward regional trade preferences already exist, and thus explain part of the proclivity toward

intra-regional trade that shows up in Table 1 and in this estimate of b. We thus conclude the

paper by using our preferred estimate of b, which comes from the gravity estimates in Part II.

They hold constant for the effects of regional trading arrangements already in existence, as well

as the effect of common languages, etc.

Table 5 gives distance in kilometers between some major world capitals. [An analogous

29

table in Frankel, Stein and Wei (1993) gives the average distance between all the pairs of

countries in our sample, by continent.] European countries tend to be both closer to each other

(and closer to the other two continents) than is the case for countries in the Western Hemisphere

or East Asia. Averaging over all countries in the sample, the mean distance between countries

on the same continent is 2896 kilometers, and on different continents is 11776 kilometers -- four

times as great. The gravity equations estimate the coefficient of the log-distance between a pair

of countries at about .56. It follows that trade between two countries on the same continent will

on average be twice as great as trade between countries on different continents, other things

equal [.56{log(11776/2896)}=.7855 and exp(.7855)=2.19].

In the algebra in Part 3 of the paper, the elasticity of demand, 0, is

given by 1/ 0. If transport costs show up fully in the

price facing the consumer, the percentage change in price

associated with being in a different continent is given by

(pnc,t/pc,t)-1 = b/(1-b) (for the case of tariffs levied on the c.i.f. value). From the data on

bilateral trade, this should be approximately equal to

0log(11776/2896) = 01.403 =

[.56 0] 1.403. Choosing again our baseline value

0=.75, our illustrative calculation suggests that the

difference between inter-continental transportation costs and

continental costs is roughly on the order of 16.4 per cent.

30

Such an estimate for b might still seem a bit high. But recent literature on spillovers and

geographic concentration suggests that the effects of proximity on stimulating production are

much greater than mere transportation costs. In the classic gravity model of world trade,

Linneman (1966) concluded that the effect of distance on trade consisted of three kinds of

effects rather than one: (i) transportation costs, (ii) the time element (involving concerns of

perishability, adaptability to market conditions, irregularities in supply, in addition to interest

costs), and (iii) "psychic" distance (which includes familiarity with laws, institutions and habits).

The .164 estimate, if taken at face value, together with our simulations suggests that the

optimal degree of preferences within a continental grouping is roughly 60 per cent, i.e., intra-

regional liberalization to 40 per cent of the level of world-wide trade barriers, in a stylized six-

country world. Only if regionalization proceeds past that point, does it enter into the zone of

negative returns to liberalization. For the more realistic 64-country world (Figure 4), negative

returns to regionalization set in as early as at 23.1 per cent preferences, and the super-natural

zone at 44.2 per cent preferences.

The last step is to try to extract from our gravity estimates of part 2 a measure of k, the

degree of preferences prevailing in existing regional trading blocs, in order to help evaluate

whether the world trading system has in fact become excessively regionalized. Our gravity

estimates in Table 2 suggest that the EC in 1990 operated to increase trade among its members

by about 50 per cent. Other parts of the world have weaker or stronger arrangements. We

have found that such FTAs as MERCOSUR and the Andean Pact actually have effects on trade

that are considerably greater (proportionally) than the EC. Let us ask the following

hypothetical question: what would be the effect on world economic welfare if the trading system

settled down to an array of regional blocs that each had the same level of preferences as the

EC?

Let the percentage effect on trade of bloc formation be represented by 0.43 It

31

is shown in Frankel, Stein and Wei (1993), that the desired

preference parameter is monotonically related to 0:

k = 0(1+t)(1- 0)/t.

Taking 0 = 0.5 from the EC estimate, 0 = 0.75, and t = .30,

the implied estimate of k is .54. In other words, EC

preferences operate to reduce trade barriers by 54 per cent

for intra-bloc trade. This parameter value lies within our

super-natural zone. It follows, within the assumptions of our

model, that if all continents followed the EC example, the

regionalization of world trade would be excessive, in the

sense that world economic welfare would be reduced relative to

the MFN norm.

* * *

The tentative conclusion of this study is that some degree of preferences along natural

continental lines, such as an Enterprise for the Americas Initiative, or enlargement of the

European Union to include EFTA and Eastern Europe, would be a good thing, but that the

formation of Free Trade Areas where the preferences approach 100 % would represent an

excessive degree of regionalization of world trade. This is especially true if the prospective

FTAs consist of entire continents. (For sub-continental regions like the Andean Pact, the

finding remains that partial PTAs can be welfare-improving, while FTAs are super-natural; but

the welfare effects are smaller than they are in the case of the continent-wide blocs) The

overall conclusion is that the world trading system is currently in danger of entering the zone of

excessive regionalization.

Any such conclusion much register some important caveats, primarily of a political

economy nature. First, although our definition of partial preferences has been a partial

32

reduction for neighbors in the tariff on all goods, in practice partial preferences usually take the

form of special consideration or outright exemption for some industries at the expense of others.

Inter-sectoral distortions and rent-seeking behavior can make this kind of partial preferences

very costly. Second is the question of the role that regional arrangements play in further

unilateral or multilateral liberalization. Assuming the ultimate goal is the achievement of free

trade among all countries, limiting the formation of blocs to geographically proximate countries

might not be the best way to go, if it leads to the permanent fragmentation of the world's trade

rather than to a process of continuous integration. The answers are not clear once we include

dynamic political economy considerations in the analysis.44

Within the terms of our model, however, the optimal path to liberalization appears to

feature a sharp departure from Article 24. It entails reducing intra-regional barriers by only 10

per cent or so. Apparently the optimal path concentrates on extending the scope of the

Preferential Trading Arrangements from two-country agreements to wider sub-continental

agreements, and then to the continental level, and then finally to the worldwide level, before

liberalizing completely within any unit. At least, such a path would under our assumptions

raise economic welfare at each step of the way.

33

References

Balassa, Bela. 1987. "Economic Integration" in The New Palgrave Dictionary of Economics,

MacMillan Press Reference Books, London: 43-47.

Baldwin, Richard. 1993. "A Domino Theory of Regionalism." NBER Working Paper No.

4465, September.

Bergsten, C.Fred. 1991. "Commentary: The Move Toward Free Trade Zones," in Policy

Implications of Trade and Currency Zones, A Symposium Sponsored by the Federal Reserve

Bank of Kansas City, Jackson Hole, Wyoming, (August): 43-57.

Bergstrand, Jeffrey. 1989. "The Generalized Gravity Equation, Monopolistic Competition, and

the Factor-Proportions Theory in International Trade," Review of Economics and Statistics 71,

no.1, Feb.: 143-152.

Bhagwati, Jagdish. 1971. "Trade-Diverting Customs Unions and Welfare Improvement: A

Clarification," Economic Journal, 63.

Bhagwati, Jagdish. 1990. "Regional Accords Be-GATT Trouble For Free Trade," Wall Street

Journal, Dec. 5.

Bhagwati, Jagdish. 1992. Regionalism and Multilateralism: An Overview. World Bank and

CEPR Conference on New Dimensions in Regional Integration, April, Washington, D.C.

34

Bhagwati, Jagdish, and T.N. Srinivasan. 1983. Lectures on International Trade, Cambridge:

MIT Press, Chapter 27.

Deardorff, Alan. 1984. "Testing Trade Theories and Predicting Trade Flows," in R.Jones and

P.Kenen, eds., Handbook of International Economics vol. I. Amsterdam, Elsevier Science

Publishers. Ch.10: 467-517.

Deardorff, Alan, and Robert Stern. 1992. "Multilateral Trade Negotiations and Preferential

Trading Arrangements," RFIE Discussion Paper No. 307, University of Michigan, July.

Forthcoming in Deardorff and Stern, eds., Analytical and Negotiating Issues in the Global

Trading System, Ann Arbor: University of Michigan Press.

de la Torre, Augusto, and Margaret Kelly, 1992, Regional Trading Arrangements, Occasional

Paper No. 93, Washington, D.C., International Monetary Fund (March).

de Melo, J., A. Panagariya, and D.Rodrik. 1992. Regional Integration: An Analytical and

Empirical Overview, World Bank and CEPR Conference on New Dimensions in Regional

Integration, April, Washington, D.C.

Edwards, Sebastian. 1993. "Latin American Economic Integration: A New Perspective on an

Old Dream," The World Economy, Basil Blackwell: 317-338.

Fieleke, Norman. 1992. "One Trading World, or Many: The Issue of Regional Trading Blocs,"

New England Economic Review, Federal Reserve Bank of Boston, May/June, 3-20.

35

Finger, J.M. 1993. "GATT's Influence on Regional Arrangements," in J. de Melo and A.

Panagariya, New Dimensions in Regional Integration (New York: Cambridge University

Press).

Foroutan, Faezeh and Lant Pritchett. 1992. "Intra-Sub-Saharan African Trade: Is It Too Little?"

World Bank, December. Forthcoming, Journal of African Studies.

Frankel, Jeffrey. 1991. "Is a Yen Bloc Forming in Pacific Asia?" in Finance and the

International Economy, The AMEX Bank Review Prize Essays, edited by R.O'Brien, Oxford

Univ. Press, UK, 1991.

Frankel, Jeffrey. 1993. "Is Japan Creating a Yen Bloc in East Asia and the Pacific?" in

Regionalism and Rivalry: Japan and the U.S. in Pacific Asia, edited by Jeffrey Frankel and

Miles Kahler, Chicago, University of Chicago Press.

Frankel, Jeffrey, Ernesto Stein, and Shang-Jin Wei. 1993. "Continental Trading Blocs: Natural

or Supernatural?" NBER Working Paper No.4588, Dec.

Frankel, Jeffrey, and Shang-Jin Wei. 1993a. "Is There A Currency Bloc in the Pacific?" in

Exchange Rates, International Trade and Monetary Policy, edited by Adrian Blundell-Wignall

and Stephen Grenville; Reserve Bank of Australia, Sydney.

Frankel, Jeffrey, and Shang-Jin Wei. 1993b. "Emerging Currency Blocs." International Center

for Monetary and Banking Studies, Geneva, Sept. 2-4. [Revised version of NBER Working

Paper 4335.] Forthcoming, The Future of the International Monetary System and Its

36

Institutions, edited by Hans Genberg, 1994.

Frankel, Jeffrey, and Shang-Jin Wei. 1994. "Yen Bloc or Dollar Bloc: Exchange Rate Policies

of the East Asian Economies", in Macroeconomic Linkages, Third Annual NBER - East Asia

Seminar on Economics, Takatoshi Ito and Anne Krueger, editors, Chicago, University of

Chicago Press.

Hamilton, Carl, and L.Alan Winters. 1992. "Opening Up International Trade in Eastern

Europe," Economic Policy (April).

Havrylyshyn, Oleg, and Lant Pritchett. 1991. "European Trade Patterns After the Transition."

Policy, Research and External Affairs Working Paper Series No. 748, August, World Bank.

Helpman, Elhanan. 1987. "Imperfect Competition and International Trade: Evidence from

Fourteen Industrial Countries," Journal of the Japanese and International Economies 1: 62-81.

Helpman, Elhanan and Paul Krugman. 1985. Market Structure and Foreign Trade, Cambridge,

MA, MIT Press.

Irwin, Douglas. 1993. "The GATT's Contribution to Economic Recovery in Post War Western

Europe." International Finance Discussion Paper No. 442, Federal Reserve Board, March.

Jackson, John. 1993. "Regional Trade Blocs and the GATT." The World Economy 121-132.

Krugman, Paul. 1980. "Scale Economies, Product Differentiation, and the Pattern of Trade,"

37

American Economic Review 70, 950-959.

Krugman, Paul. 1991a. "Is Bilateralism Bad?" in E.Helpman and A.Razin, eds., International

Trade and Trade Policy. Cambridge, MA, MIT Press.

Krugman, Paul. 1991b. "The Move Toward Free Trade Zones," in Policy Implications of

Trade and Currency Zones, A Symposium Sponsored by the Federal Reserve Bank of Kansas

City, Jackson Hole, Wyoming, (August): 7-42.

Krugman, Paul. 1992. Regionalism vs. Multilateralism: Analytical Notes, World Bank and

CEPR Conference on New Dimensions in Regional Integration, April, Washington, D.C.

Lawrence, Robert. 1991. "Emerging Regional Arrangements: Building Blocks or Stumbling

Blocks?" in Finance and the International Economy, The AMEX Bank Review Prize Essays,

edited by R.O'Brien. United Kingdom: Oxford University Press.

Linneman, Hans. 1966. An Econometric Study of International Trade Flows, North-Holland,

Amsterdam.

Meade, James. 1955. The Theory of Customs Unions. Amsterdam: North-Holland.

Saxonhouse, Gary. 1993. "Pricing Strategies and Trading Blocks in East Asia," in

Regionalism and Rivalry: Japan and the U.S. in Pacific Asia, edited by Jeffrey Frankel and

Miles Kahler, Chicago: University of Chicago Press.

38

Srinivasan, T.N., 1992, "'Regionalism vs. Multilateralism: Analytical Notes' Comment,"

World Bank and CEPR Conference on New Dimensions in Regional Integration, April,

Washington, D.C.

Stein, Ernesto, and Jeffrey Frankel, 1993, "Transport Costs and the Welfare Effects of Free

Trade Agreements," February; Chapter in Ph.D. thesis, E.Stein, University of California,

Berkeley.

Summers, Lawrence. 1991. "Regionalism and the World Trading System," in Policy

Implications of Trade and Currency Zones, A Symposium Sponsored by the Federal Reserve

Bank of Kansas City, Jackson Hole, Wyoming, (August): 295-302.

Wang, Zhen Kun, and L.Alan Winters, 1991, "The Trading Potential of Eastern Europe,"

Centre for Economic Policy Research Discussion Paper No. 610, November, London, UK.

Wonnacott, Paul, and Mark Lutz, 1989. "Is There a Case for Free Trade Areas?" Chapter 2 in

Jeffrey Schott, ed., Free Trade Areas and U.S. Trade Policy, Washington, D.C.: Institute for

International Economics, 59-84.

* * *

Notes . The Latin American Integration Association (LAIA) replaced the LAFTA in 1980. De la Torre and Kelly (1992) and

Fieleke (1992) chronicle the lapses between proclamation and practice in these cases [and others, such as ECOWAS in West Africa], in their surveys of the post-war history of regional trading arrangements. Edwards, 1993, reviews the history of regional economic arrangements in Latin America.

39

. The CACM, comprising six Central American countries, was strengthened in 1991. Central American countries have

recently joined Caribbean countries (who have their own arrangement, CARICOM), in asking North America for "parity"

with Mexico of access to NAFTA in at least some of their exports.

. Consistent with the Enterprise for the Americas Initiative proposed by the Bush Administration in June 1990.

. Modern trade theory has come up with a number of circumstances in which unilateral subsidy or other intervention by

one country's government is capable of making that country better off (e.g., certain technological spillovers, and strategic

industries). But the models do not undermine standard "free-trade policy," which holds that a world in which governments

cooperatively agree to limit subsidies or tariffs is better than a world where all are left free to undertake them. To the

contrary, the new models usually tend to strengthen the case for multilateral agreements, though this is not explicitly

recognized as often as it might be. (These models' conclusions also tend to be very sensitive to imperfect knowledge on the

part of governments, or vulnerability to political influence by interest groups.)

. Bhagwati (1992), Deardorff and Stern (1992), and de Melo, Panagariya, and Rodrik (1992) review the literature.

Fieleke (1992) is a useful non-technical review of regional trading arrangements.

. Jackson (1993, p.123), for example, has suggested that the goal of the Article 24 exception to the MFN principle is that

FTAs would be trade-creating rather than trade-diverting.

. See Lawrence (1991). A good argument for the NAFTA, for example, is that it locks into place trade liberalization

that Mexico had undertaken anyway, but that future political forces in Mexico might seek to change. Many considered this

argument to be as important as the economic gains from the NAFTA provisions in their own right [or as other considerations

such as immigration or U.S.-Mexican political relations].

40

. A later contribution, Krugman (1992), dropped the assumption of optimal or endogenous tariffs. The conclusions were

similar, which showed that another effect was at work in addition to the monopoly-power effect, as explained below.

. In what the Economist called "the shootout at Jackson Hole," Summers (1991) agreed with Krugman that natural blocs

were likely to be beneficial, while Bergsten (1991) was on the other side. [It should be noted that the idea of proximity as a

desideratum for successful FTAs, on the grounds that it would minimize the amount of trade diversion, was not entirely new

with Krugman. (See Balassa 1987, p.44, and Wonnacott and Lutz 1989).] The leading opponent is Bhagwati (1992), whose

reaction to reports from Jackson Hole was: "The prescription is sufficiently strange and hard to defend for me to wonder

whether these distinguished economists truly expressed these views" (footnote 8).

.. In this paper we emphasize the sub-regional groupings. A companion paper emphasizes the continents: Frankel, Stein

and Wei (1993).

.. The term was introduced in Frankel (1993).

.. The United States initially opposed discriminatory tariff policies such as the British Commonwealth preferences in the

founding of the GATT, but dropped its opposition in the 1950s in the context of European integration, the GATT rules

notwithstanding. Irwin (1993) and Finger (1993) review the history.

.. Saxonhouse (1993) considers this question.

.. The results of one extensive early project along these lines were reported in Linneman (1967). Foundations for the

gravity model are offered in papers surveyed by Deardorff (1984, pp.503-06) and Wang and Winters (1992), such as

41

Linnemann (1966).

.. The focus of these papers was on potential Eastern European trade patterns. The Winters papers report statistically

significant within-region biases to the following groupings: EC, Latin America, ASEAN, former British colonies, GSP, and

EC preferences under the Lome convention. Havrylyshyn and Pritchett (1991) report significant effects for the EC, LAFTA

and CACM.

.. The list of countries, and regional groupings, is given in an Appendix to Frankel (1994) and the Frankel and Wei

papers. The Andes Pact grouping consists of Bolivia, Peru, Ecuador, Colombia and Venezuela, as well as Chile.

.. In other words, a rich country will trade more with another rich country than with a poor country, even if the latter

has a larger population so that its total GNP is as large as the others. This property of some modern trade theories directly

contradicts the classical Heckscher-Ohlin theory of comparative advantage based on differences in factor endowments.

Bergstrand (1989) includes both imperfect-competition effects and factor-proportions effects in his econometric gravity model.

.. Reported as Table 2 in the original conference paper version [available as CIDER Working Paper No. C94-034], but

omitted here to save space.

.. In some cases, e.g., the EC, these results confirm what one might have guessed from looking simply at intra-regional

trade shares, as in Table 1. But in other cases, e.g., EFTA and MERCOSUR, the corrections of the gravity model make a

big difference. [Typical of many studies using simple intra-regional shares, Wonnacott and Lutz, p.76, show increases in

intra-regional trade attributed to the formation of the EC, EFTA and the Andes Pact, and not to LAFTA or ASEAN.]

.. Tables 4 and 4a in the original conference paper version [available as CIDER Working Paper No. C94-034], omitted

42

here to save space.

.. Havrylyshyn and Pritchett (1991) found that three languages are significant in the gravity model -- Portuguese,

Spanish and English, in decreasing order of magnitude. In a study of poor countries, Foroutan and Pritchett (1992) find that

French, Spanish and English are statistically significant.

.. The results are reported in Table 5a of the original conference paper version and in Table 5b for the case of

manufactured goods [available as CIDER Working Paper No. C94-034]. The language coefficient is not statistically

significant when the test is run as in Table 5b, where the inclusion of five individual major languages create multicollinearity

with the general language term. But the coefficient is significant for half the years when the analogous test is run on

aggregate trade (Table 6 in Frankel and Wei, 1993b) and is highly significant for all years when the coefficient is constrained

to be the same for different languages (Table 1 in Frankel and Wei, 1993a).

.. Taiwan-China trade does not appear in the statistics, because it is officially non-existent. Much of it goes through

Hong Kong, and is thus counted twice. An attempt to correct for this factor eliminates the extra effect of the Chinese

language term (Table 3 of Frankel and Wei, 1993a).

.. The notion of transportation costs should probably be understood as transactions costs, encompassing not just physical

transportation of goods but also costs of communications and the idea that countries tend to have a better understanding of

their neighbors and their institutions.

.. The c.i.f.-based assumption is pursued in another working paper.

.. In Stein and Frankel (1993), we also examine implications of FTAs for trade diversion and trade creation, which

43

contributes some intuition to the welfare results.

.. Krugman (1991a) considers for his simulations three different values for the elasticity of substitution: 2, 4 and 10.

Since the elasticity of substitution is equal to 1/(1-è), the middle value of 4 is equivalent to our value of è=0.75.

.. In Figure 1 the level of welfare is normalized to be 1 in the case of a single bloc, i.e., the case of worldwide free

trade.

.. As in Krugman (1992).

.. Figure 3 in Stein and Frankel (1993) represents the effects of agreements on welfare for è=0.85 and t=0.35. In this

case, the effect of the formation of natural trading blocs on welfare is substantial, even for low levels of intercontinental

transportation costs. Indeed, for these parameter values, blocs are welfare-improving even when b=0. [The intuitive

explanation is that residents consume so much of the home good, that it is a net gain to realign correctly the relative price of a

neighbor's good in terms of the home good, even though this distorts the relative price of the neighbor's good in terms of all

goods produced elsewhere in the world.] Thus Krugman's idea that the consolidation of six blocs into three in the absence of

transportation costs is bad depends on the values of the parameters t and è.

.. For each set of parameter values (transport cost and è) welfare is normalized to be 1 under free trade in the figure.

.. This follows from the fact that the welfare functions are strictly concave to the origin so, in general, the maximization

problem will have an interior solution.

.. Bhagwati (1992) discusses possible reasons for the inclusion of Article 24 in the GATT.

44

.. Figure 4. Figure 4b in that working paper also does it for the 64-country case. Figure 4 in the CIDER working

paper C94-034 version of this paper does the 64-country case with tariffs levied on the f.o.b. price [and shows b running all

the way up to 1].

.. Note that the "super-natural" bloc area does not always exist. It depends on the parameter values.

.. The levels of welfare corresponding to the FTA extreme and the optimally-chosen PTA are graphed, for the range of

values of b, in Figure 5 of Frankel, Stein and Wei (1993).

.. We have also tried a simulation where the world consists of 3 continents, each of them containing 12 countries. The

results are similar to those shown here. But carving up each continent into two blocs of six countries each, when b = .2,

turns out to be the welfare minimum: not only worse than MFN or smaller FTAs, but worse also than continental FTAs.

.. These results do not allow for the fact that transport costs between potential sub-regional Free Trade Areas, such as

North and South America are greater than between countries within the same sub-region. But estimates similar to those made

here for inter-continental transportation costs would be smaller in the case of North-to-South America costs, making

regionalization on such a scale more likely to be excessive than on the Asia-Americas-Europe scale.

.. Deardorff and Stern (1992, p.17-20) suggest as much.

.. Table 36 from Review of Maritime Transport 1990, UNCTAD, U.N.: New York, 1991.

.. Krugman (1991) and Summers (1991), for example, use simple calculations to infer roughly the importance of distance

45

in determining trading patterns, without explicitly distinguishing the effect of existing trade preferences.

.. Foroutan and Pritchett (1992) find that the 19 African countries in their sample trade more with each other than the

other gravity variables would be predict, though the bloc effect is only of borderline significance.

.. The coefficient in the gravity equation is actually the log of (1+ Error! Main Document Only.).

.. Political economy considerations like those mentioned in the introduction -- a country that joins an FTA may then

experience an increase in political support for further steps toward liberalization -- are modelled by Baldwin (1993), and also

in a preliminary way in Wei and Frankel (1993).